Feedforward controller of rotating electrical machines

The control of electrical machines is almost systematically based on the principle of nested loops. In the very common case of speed control, the internal loop must impose the machine torque; it receives a setpoint generated by the speed loop. In the case of position control, the speed loop receives a setpoint generated by an external loop, the position loop. Thus, controlling the speed and position of an electric machine requires precise control of the torque it supplies.

A great deal of work has been carried out with the aim of achieving ever greater torque dynamics, ever lower steady-state oscillations... Among the control principles that have emerged, predictive control offers many advantages, such as simplicity of concept, intuitive settings and ease of implementation. However, it is significantly more computationally intensive than its competitors. As a result, its implementation has only recently become possible, thanks to the availability of fast and inexpensive calculation units.

The concept of predictive control applied to converter-machine assemblies is presented in this dossier. The example of a permanent magnet synchronous machine associated with a three-phase two-level inverter is then used to show, step by step, how this control principle can be applied to a given application. Finally, examples of systems and control variants are studied to give an overview of the possibilities offered by predictive control of rotating machines.

Modeling systems consisting of a power electronics converter associated with a rotating machine brings to light continuous quantities (i.e. those that cannot be discontinuous), such as the current in the machine windings or the voltages across capacitors. When power electronics components are used, the switching times of diodes and transistors are several orders of magnitude shorter than the time constants of other system elements. As the transient regimes that occur when power electronics components are switched may not be taken into account in the model used by the control system, these components can be assumed to have only two states (on or off). As each power electronics component can have a finite number of states, it appears that the number of possible configurations of the complete converter is finite. Continuous variables are therefore unsatisfactory for describing converter-machine assemblies. It therefore seems natural to use discrete variables (i.e. variables that can take on a finite number of values) to model the different states of power electronics components. The systems under consideration (converter-machine combinations) are therefore modeled using both continuous and discrete variables.

The approach presented aims to control a power converter-rotating machine...